| From: | Rudi Chiarito |
| Date: | 25 Sep 2000 at 16:16:43 |
| Subject: | Re: AMIOPEN: XML and Java, a uncharted territory? |
On Mon, Sep 25, 2000 at 02:19:04PM +0200, Ole Friis Ostergaard wrote:
> Hmmm, so memory fragmentation may become a problem, since there is no
> compaction taking place. Surely this allows for good speed (since no
I don't know if no compaction takes place. The documentation doesn't
mention it but it doesn't mean that e.g. some simple form of compaction
or an alternative to compaction isn't performed.
Is one going to use a lot of different objects of totally different
sizes? That's the only scenario in which I can imagine fragmentation
problems. If the objects you allocate most often are of a certain size
(or two sizes or three, you know what I mean), i.e. if there's a pattern
in allocations, the memory could easily be "recycled". Especially if the
underlying system distributes allocations in pools, according to their
size. Other optimisations involving pools are possible.
I recall from discussions held with Tao many, many months ago that memory
management (not just under Java, but under Elate in general) is quite
smart. One of the first things we investigated was how the memory
allocation scheme(s) work compared to the one used by the Classic
AmigaOS.